Comments

Health Affairs encourages readers to engage in discussion via comments on our Web site.

• To RESPOND to a particular article: Click on the link "Submit a response to this article" in the box at the top right-hand corner of the article.
• To READ responses to a particular article: Click on the link "View responses" in the box at the top right-hand corner of the article.

Comments to:

Aaron S. Kesselheim, Michael A. Fischer, and Jerry Avorn Extensions Of Intellectual Property Rights And Delayed Adoption Of Generic Drugs: Effects On Medicaid Spending Health Affairs, November/December 2006; 25(6): 1637-1647. [Abstract] [Full Text] [Figures Only] [PDF] [Online Appendices][Erratum] [Reprints & Permissions]

Comments:Submit a response to this article

Comments published:

Potential Bias and Confounding Effects: Challenges in the Analysis of Prescription Data

Kennon R. Copeland, Stephen J. Boccuzzi   ( 30 May 2007 )

Potential Bias and Confounding Effects: Challenges in the Analysis of Prescription Data 30 May 2007
Kennon R. Copeland, Senior Director - Statistical Services IMS Health, Stephen J. Boccuzzi Send comment to journal: Re: Potential Bias and Confounding Effects: Challenges in the Analysis of Prescription Data kcopeland{at}us.imshealth.com Kennon R. Copeland, et al.	[Editor's note: This lengthy eLetter has not been peer reviewed. A shorter version of it, along with the authors' rebuttal, will appear in the print edition of Health Affairs (July/August 2007).] The study by Kesselheim et al. utilized Medicaid prescription drug data and acquisition costs for three drugs from first quarter 2000 through fourth quarter 2004. The authors claim that “[the] delay in availability, elevated prices, and slow uptake of generic alternatives for these three drugs alone cost Medicaid $1.5 billion in 2000-2004.” The authors also claim that their findings suggest that overall drug expenditure savings would be billions of dollars for non-Medicaid payers. While important economic issues are addressed, the study suffers from several biases and limitations related to research methodology and analytic approach. These weaknesses affect the validity and generalizability of the findings, including extrapolation of drug savings to all payer segments and the economic impact related to the three therapeutic agents within the Medicaid population. Definition of Product Costs for Study Period The authors base their savings calculations on the lowest observed generic average cost across the study period in conjunction with the actual observed brand average cost for each quarter. By definition, this approach produces cost savings. The rationale for using the lowest generic cost as the basis for cost savings in all prior quarters while assuming that the brand average cost would be unaffected is unclear. In addition, the assumption related to the decrease in average costs across time and its affect of cost trends needs more clarity. The lowest observed generic average cost occurs in the last quarter of the study period for the study drugs of interest (N.B. lowest OTC costs for omeprazole occurred in an earlier quarter), and is noticeably lower than the average cost seen for any prior quarter -- 8% to 13% lower than the average cost in the immediately prior quarter and 25% to 120% lower than the highest observed average price. Many market factors preclude the reader from assuming that the most recent cost is an appropriate representation of the product’s price throughout the study. The authors state, “For each study drug, the cost to Medicaid fell once the six-month generic exclusivity period ended and multiple generic products became available. This decrease in costs was 22 percent for omeprazole, 24 percent for amoxicillin/clavulanate, and 55 percent for metformin.” However, the decrease in costs did not occur immediately, as would have been assumed. According to the authors’ exhibits, the decreased costs were observed in the last quarter of the study period -- two to three years after the end of the generic exclusivity period. For the six-month period immediately following the six-month generic exclusivity period, the decrease in costs were: 2% for amoxicillin, 4% for omeprazole, and 14% for metformin. A more reasonable approach would be to use the actual average generic costs or to have developed a time-variant predicted average cost based upon the actual average cost trends and brand-generic cost ratio trends over time. These approaches would be less likely to result in a misrepresentation of potential savings. Use of OTC Cost for Prescription Omeprazole The authors calculate potential savings utilizing the lowest average cost for OTC omeprazole rather than the lowest average cost for generic omeprazole. Use of the OTC version of omeprazole as the lowest average cost resulted in a basis price of $0.55, 79% less than what would have been expected had the authors used generic omeprazole ($2.60). This approach drives the majority of potential savings for omeprazole. Further, OTC omeprazole did not become available until four quarters after the introduction of generic omeprazole, which suggests the need for a temporal adjustment related to market access. Alternative approaches could include using the actual average costs for generic omeprazole or a predicted average cost based upon the actual average cost trends over time or determining the proportion of OTC and generic omeprazole utilization and using the average costs based on the cost trends over time for these two drug proportions. Estimation of Potential Savings Associated with Earlier Generic Availability The authors appear to have an error in the Appendix 2 formula that was used in determining potential savings associated with earlier generic availability. It is not clear whether the mistake was merely in documentation, or whether the mistake affected the analysis used to generate the results. [Editor's note: This error has been corrected.] The allocation of brand and generic utilization could have been more accurately derived by applying the proportion of brand or generic utilization, to total units dispensed, not to the corresponding brand and generic units dispensed, so as to yield brand/generic utilizations that sum to the total utilization. The formula used in the analysis would therefore result in an underestimate of utilization and in incorrect brand and generic proportions. The use of lowest average generic cost also does not appear appropriate. Calculation of Per Unit Cost The method used to derive per unit costs for each drug of interest is not clear, as each agent comes in various strengths, and the per unit cost appears to be based upon number of pills. Drug prices per pill can vary by strength and are not generally linear; thus, the per unit cost would be affected by the distribution and strength of each drug. The authors’ assumption appears to be that the utilization by strength is the same for both brand and generic drugs. A more appropriate approach would involve calculating a per unit cost for each drug strength and then deriving a weighted average for the drug across strengths adjusting for differences in average costs by strength. Adjustment for Rebates This study applies one set of rebate adjustments for brand products and another for generic products, using average rebates from a 2003 CBO analysis. However, drug rebates vary widely from the average rebate for both branded and generic agents. The authors also utilized typical discounts from average wholesale price (AWP) and typical relationships between average manufacturer price (AMP) and AWP to derive estimated rebate amounts. A better estimate of the specific rebate potential for each drug of interest could have involved incorporating the utilized volume of the product vs. national share with sensitivity analysis to account for the impact of rebate variability and discounts on stated potential savings.